China’s rising hydropower demand challenges water sector

Demand for hydropower is increasing, yet the water footprints (WFs) of reservoirs and hydropower, and their contributions to water scarcity, are poorly understood. Here, we calculate reservoir WFs (freshwater that evaporates from reservoirs) and hydropower WFs (the WF of hydroelectricity) in China based on data from 875 representative reservoirs (209 with power plants). In 2010, the reservoir WF totaled 27.9 × 109 m3 (Gm3), or 22% of China’s total water consumption. Ignoring the reservoir WF seriously underestimates human water appropriation. The reservoir WF associated with industrial, domestic and agricultural WFs caused water scarcity in 6 of the 10 major Chinese river basins from 2 to 12 months annually. The hydropower WF was 6.6 Gm3 yr−1 or 3.6 m3 of water to produce a GJ (109 J) of electricity. Hydropower is a water intensive energy carrier. As a response to global climate change, the Chinese government has promoted a further increase in hydropower energy by 70% by 2020 compared to 2012. This energy policy imposes pressure on available freshwater resources and increases water scarcity. The water-energy nexus requires strategic and coordinated implementations of hydropower development among geographical regions, as well as trade-off analysis between rising energy demand and water use sustainability.


Results
Reservoir WF. The Chinese reservoir WF totaled 27.9 × 10 9 m 3 (Gm 3 ) in 2010, with values ranging from 0.7 Gm 3 for the Northwest rivers basin to 8.0 Gm 3 for the Yangtze River basin ( Table 1). The reservoir WF accounted for 22% of the total blue water WF of China (Tables S1, S2); this proportion ranged from 5% for the Northwest rivers basin to 57% for the Southeast rivers basin (Table S2). Hence, neglecting reservoir WF seriously underestimates the blue water WF.
The reservoir WFs varied widely among the basins (Fig. 1) and months ( Table 1). The Yangtze and Zhujiang river basins have the highest reservoir WF; they account for 46% of the total national reservoir WF. One explanation is that about 45,200 dams have been built throughout the Yangtze River basin, which amounts to about 51% of the Chinese total 20 . The Southwest and Northwest river basins have the lowest reservoir WF; they account for about 6% of the national total. In the Northwest river basin, precipitation is low and the climate and topography are unsuitable for the construction of large dams. The WF from May to September accounts for more than half of the annual WF in all basins, except in the Southwest rivers basin ( Table 1). The Huang, Huai, Hai, and Yangtze river basins have large reservoirs (WF > 10 × 10 6 m 3 ) that coexist with small reservoirs (WF < 10 × 10 6 m 3 ), whereas large reservoirs dominate the Southeast rivers basin.
Hydroelectric WF. China's hydroelectric WF totaled 6.6 Gm 3 yr −1 in 2010. This was about 24% of the reservoir WF. We also calculated the product water footprint (PWF; the WF per unit of goods or services, including hydropower generation 16 ). PWF varied widely among the plants (Fig. 3). The largest hydropower plant, the Three Gorges Dam on the Yangtze River, had a product WF of 2.0 m 3 GJ −1 ; the second-largest plant, the Gezhouba Dam on the Yangtze River, had a PWF of only 0.6 m 3 GJ −1 . There is a general trend of relative lower PWF at upstream and larger PWF at downstream within a river basin, e.g. for the Yangtze and Zhujiang river basins (Fig. 3). Table S5 summarizes the hydroelectric PWF for the 209 power plants. By weighing the WF of the representative hydropower plants by their power generation, we calculated a weighted average hydroelectric PWF of 3.6 m 3 GJ −1 . Hydroelectric PWF varied widely, from 0.001 m 3 GJ −1 for the Hongyi plant to 4234 m 3 GJ −1 for the Zhanggang plant, both in the Yangtze river basin. This range is much wider than previous IPCC estimates of between 0.011 and 58 m 3 GJ −1 (or 0.04 to 209 m 3 MWh −1 ) 11 . The IPCC summarized the range based on two data points from four literature sources 17 . Comparatively speaking, our results are more solid and specific by including 209 officially surveyed hydropower plants in China.

Discussion
The reservoir WF is seldom accounted for when assessing the human appropriation of water resources or the pros and cons of constructing reservoirs to provide hydroelectricity. We provided a spatially explicit assessment of the reservoir and hydropower WFs by considering 875 representative reservoirs and 209 hydropower plants in different locations, and demonstrated that accounting for these WFs is an important consideration when evaluating the environmental sustainability of a reservoir or of hydropower as an energy source. We therefore recommend that the WF should be assessed in any sustainability evaluation for new reservoirs, as well as in the evaluation of existing reservoirs, so that the consequences of the WF on downstream environmental flows and other water users can be evaluated. About one-fourth of the world's reservoirs with a dam higher than 15 m provide multiple services, and more than 40% of the 8689 reservoirs that provide hydropower are also used for other purposes 17 . For multi-purpose reservoirs, it is more logical to "share the burden of water consumption" among the different beneficiaries 17 , as we have done in the present study. Annual and monthly water scarcity assessment for the 10 Chinese river basins. ADI represents the agricultural, domestic and industrial sector (i.e., excludes the reservoir WF). Note that the y-axis scale differs from graph to graph. Different color tones indicate different levels of water scarcity: red for severe, orange for significant, yellow for moderate and blue for low. In the graph, for each river basin, the color tones associated with the bars indicate the same severity of water scarcity. For example, when the blue WF (the grey and black bars) reaches a red zone, severe water scarcity occurs in that month for the basin. The PWF of 3.6 m 3 GJ −1 calculated here is lower than several previously reported hydroelectric PWFs: e.g. 68 m 3 GJ −1 by Mekonnen and Hoekstra 13 and 22 m 3 GJ −1 by Gerbens-Leenes et al. 12 for the global average, 6.1 m 3 GJ −1 for New Zealand 15 and 19 m 3 GJ −1 for the United States 25 (Table 2). Although the reservoirs of many power plants are also used for other purposes (e.g., irrigation supply), the WF was attributed only to hydropower in all previous studies except one 26 . Many of these studies were based on a limited number of power plants, and may therefore contain large errors in the hydroelectric WF. Mekonnen and Hoekstra 13 considered only 35 of the world's 8689 plants 17 , which accounted for 8% of global electricity generation 17 . Moreover, the selected plants were not broadly representative, and included only two Chinese plants and no plants in the United States. In addition, Gerbens-Leenes et al. 12 and Mekonnen and Hoekstra 13 did not use recent data on electricity generation. Those studies are pioneer work but results are insufficiently representative. We allocated the total WF among the different outputs of the reservoirs, and to the best of our knowledge, we are the first to obtain results and analyze the variances based on a large number of reservoirs (i.e. 875 for China) and hydropower plants (i.e. 209 for China) and to demonstrate the spatial distribution of WF of reservoirs and hydropower.
The Chinese national average hydroelectric PWF of 3.6 m 3 GJ −1 is higher than that of most other technologies, which typically have maximum values of 1.1 to 1.4 m 3 GJ −1 (4 to 5 m 3 MWh −1 ) 27 . At present, the most important primary energy carriers include crude oil, coal, natural gas, uranium, hydroelectric power, solar and wind energy 28 . The PWF of wind energy and underground uranium mining is negligible 29 . The water footprint of electricity from solar energy, coal-fired and nuclear thermal energy is generally far below 1.0 m 3 GJ −1 14,29 (Table 3). Thus, hydropower is not an efficient solution to energy supply from a water consumption perspective. As a large water consumer, the hydropower sector has triggered great concern 27 since it may result in reputational and financial risks, in particularly in regions where freshwater is scarce.
China covers a large geographic area, with a wide and complex variation in climatic conditions, thus we expect no correlation between latitude and WF; our results ( Fig. S1) confirm this hypothesis. Evaporation is a key parameter in influencing the WF. Evaporation varies from 748 mm yr −1 for the Ankang reservoir (Yangtze River basin) to 1646 mm yr −1 for the Yantan reservoir (Zhujiang River basin) (Fig. 4B). Most reservoirs had an evaporation rate between 900 and 1500 mm yr −1 , but tropical reservoirs generally had higher evaporation rates than temperate-zone reservoirs (Fig. 4B). However, different evaporation rates cannot explain the large variation in PWF.
We found that the large variation of the PWF was mainly determined by the reservoir area per unit of installed hydroelectric capacity (α ). The value of α averaged 9.9 ha MW −1 , with a minimum of 0.004 ha MW −1 (Dayingjiang plant, Southwest rivers basin) and a maximum of 5280 ha MW −1 (Zhanggang plant, Yangtze River basin). Plants with a relatively large α generally had a larger PWF than those with a small α (Fig. 5). We found a linear relationship between PWF and α (Fig. 5). This linear relationship was much stronger for plants with hydropower as their main purpose than for those with power as their secondary purpose. Plants with a high installed hydroelectric capacity also had a strong linear relationship with PWF ( Fig. 5), probably because they were often constructed primarily to provide hydropower. Hence, the reservoir surface area also plays an important role in the size of the hydropower WF 13,17 . To our best knowledge, very few studies reported on the relationship between the PWF of electricity, the reservoir area and the installed hydroelectric capacity except for the study of Mekonnen and Hoekstra 13 . The large reservoir area is not exclusively due to the need for power production, as the reservoir may be used for multiple purposes (e.g., irrigation supply or flood control), or originates from a natural lake before there was a reservoir 17 .
The generally lower PWF upstream of the major river basins can largely be explained by the close relation between the PWF and α . In China, several large rivers, such as the Yangtze River, originate from high-altitude western regions. The upstream areas are often characterized by steep mountains, while  17 . Note: In this table, the hydroelectric PWF is presented based on the same definition (i.e., the evaporative water consumption for each unit of hydropower generation). Values in m 3 MWh −1 were calculated by multiplying the values in m 3 GJ −1 by 3.6 (the conversion factor).
downstream areas are dominated by flat plains. These topographical differences make upstream areas more suitable to construct relatively small-area reservoirs. Water scarcity depends on many factors, such as blue water consumption, water management and precipitation . In this paper, water consumption is reflected by the concept of the water footprint, which refers to water consumption in different sectors. Water resources come directly or indirectly from precipitation. Water management is a factor that can influence water scarcity, but management is also included in the estimation of water footprint. For example, irrigated agriculture differs from rainfed agriculture, generating a blue WF. The sustainability of monthly blue water consumption is assessed by comparing blue WF with blue water availability (the difference between blue water resources and environmental flow requirements). The general idea of the monthly blue water scarcity assessment is to judge whether the human induced blue WF is beyond or below blue water availability in a month, and such an assessment approach is recommended by the Water Footprint Network 16 .
To decrease the human impact on climate change, the Chinese government promotes the use of renewable energy, in particular hydropower 20 . The reservoir WF accounts for over one-fifth of the total human water appropriation in China. A 70% increase in installed hydropower capacity planned in the national energy policy 20 will further increase the hydropower WF, as well as the reservoir WF, unavoidably resulting in more competitive water consumption between energy supply and other purposes, such as water for food production. There may be large trade-offs between food security and energy security in China given its already serious water scarcity. The reservoir WF may reduce the availability of water for food production, but at the same time, the regulating role of reservoirs may benefit food production. After a reservoir is constructed, it can restore flood. As a result, food production can benefit from reduced flood damage to agricultural fields and from increased availability of irrigation water in the reservoir. The complex relation makes the trade-off assessment more difficult, although such an assessment is beyond the scope of the current paper.
China's hydropower resources are mainly concentrated in the western regions 30 , where the PWF is generally low; but energy demand is dominant in the populous eastern regions with a generally high PWF. From a water conservation point of view, eastern China should not further expand its capacity in hydropower. Future reservoir construction in eastern China should avoid aiming at hydroelectric production. Hydropower capacities can be increased strategically and coordinated by the national government, with mutual agreement with provincial governments. On the other hand, many ongoing  hydropower projects are located in the ecologically important but vulnerable western regions 30,31 , which are sources of many large rivers, including the Yangtze River and the Mekong River that is shared by six countries. There is a need for strategic and coordinated implementation of hydropower development among different geographical regions. China's future hydropower development must take into account nature conservation, biodiversity protection 30 , as well as ecological resilience locally and for downstream regions 31 . Furthermore, hydropower development should take the impacts of hydropower on water use sustainability and global climate change into account in addition to the raising energy demand. Our study has several limitations. First, reservoirs also have beneficial impacts. They store water for agriculture during dry months, thereby alleviating drought and increasing food production. Moreover, they reduce the risks of devastating floods in downstream regions 19 . With proper management, reservoirs can increase water availability during dry seasons and prevent floods in wet seasons. Unfortunately, these benefits cannot be handled by the present simplistic analytical approach. A broader conceptual framework that accounts for trade-offs among different water uses is still needed, but is beyond the scope of our study.
Second, we calculated the reservoir WF by accounting for the total evaporation. It is reasonable to note that before the reservoir was created, there was also evaporation from the area 27 . However, evaporation from the original flowing river is likely to be considerably less than that from the reservoir, since the reservoir area is generally much larger than the river's original area 27 . We therefore argue that the full reservoir evaporation (without subtracting evaporation from the original rivers) should be considered when quantifying the water consumption that can be associated with a specific human purpose, such as electricity generation. In addition, after a reservoir is constructed, precipitation can be completely turned into runoff over the reservoir surface; hence, more blue water is available. However, it remains a shortcoming for this study that cannot quantify the change of blue water resources caused by the reservoir construction. Furthermore, we only included the functional reservoir evaporation, and excluded the supply-chain WF of hydroelectric generation; that is, we excluded the WF of the materials used in the construction, operation, and maintenance of the site, although this is negligible compared to the functional WF 27,32 .
Third, the water that evaporates from a reservoir can be considered "lost" because, unless it falls as precipitation close to its source, it cannot be reused in the same catchment. However, some of this water may be recaptured within the same catchment, particularly in areas with steep mountains that create orographic precipitation. Reservoirs also have a temporally dynamic water surface area. As a result of annual and seasonal fluctuations in the water volume, the difference between the minimum and maximum reservoir area will also vary. Reported areas generally refer to the maximum, and this probably leads to some overestimation of annual evaporation.
Finally, using economic values to allocate water consumption to hydroelectric power generation does not tell the entire story, since the "social value" of water (e.g., for irrigation or recreation) might be higher than its economic price. In addition, there is an omission of including opportunity costs in the methodology. Consumption of different water users is inevitable and one has to ensure that consumption takes place in those sectors where there is more economic value per unit of water. Hydropower is the lowest cost energy option and its generation provides extensive benefits to major economic benefits and contributes to growth and improvements in social well being. Thus, although our study provides a strong first attempt to demonstrate the importance of reservoir and hydroelectric WFs, it will be necessary to develop a more sophisticated methodology for estimating water consumption from multi-purpose reservoirs to obtain better estimates that can be used to support reservoir management 27 .

Methods
Reservoir WF. The WF of a product or service is defined as the volume of freshwater used to produce that good at the place where it was actually produced 16 . A WF consists of three components: the green WF, the blue WF, and the grey WF 16 . The green WF refers to soil water from precipitation that evaporates during the production of a good or service. The blue WF refers to surface and groundwater consumed during the production of a good or service. The grey WF is related to water pollution and is defined as the amount of water needed to dilute pollutants discharged into natural water systems to the extent that the quality of the ambient water remains better than the required water quality standards 16 . Based on these definitions, hydroelectric power only generates a blue WF.
We selected 875 representative reservoirs in China's 10 river basins (Fig. 5A): the Songhuajiang, Liaohe, Haihe, Huanghe (Yellow River), Huaihe, Changjiang (Yangtze River), Southeast rivers, Zhujiang, Southwest rivers, and Northwest rivers basins. These representative reservoirs accounted for 76% of the total reservoir volume in China. There are two main criteria to select these reservoirs: (1) A good coverage of large-and medium-sized reservoirs; and (2) data availability. Among these reservoirs, 775 are included in the GRanD database 33 , which contains information on the location, main use, dam height, storage capacity, surface area and the name of the basin where the dam is located. We selected the other 100 reservoirs as follows: there are 185 large and medium-sized hydroelectric power plants in China 34 , of which 55 plants are included in the GRanD database 33 . Among the remaining 130 plants, data on either the reservoir volume or height are available for 100 plants (see the data sources in SI Appendix I). We calculated the reservoir area based on one of the following two types of regression functions 33 : when both the reservoir volume and height were available, we used a regression function that related the reservoir area to these two variables; when only the reservoir volume was available, we used another regression function that related the reservoir area to the reservoir volume 33 . Table S1 provides the data for the 875 reservoirs.
We first calculated the annual WF of a representative reservoir i in river basin j, F (m 3 yr −1 ), which was equivalent to the total reservoir evaporation on an annual basis, by multiplying the annual water evaporation by the surface area: where, E (mm yr −1 ) is the water evaporation and A (ha) is the surface area of reservoir i in river basin j. Since 1 mm is equal to 10 m 3 ha −1 , the factor 10 is used to convert mm into m 3 ha −1 . We used the measured reservoir evaporation from literature for 69 reservoirs (Tables S6 and S7) to estimate the WF of these reservoirs. These reservoirs accounted for 62% of the total power generation and 40% of the total storage volume of the 875 reservoirs. When measured data were unavailable, we calculated the annual evaporation E (mm yr −1 ) from the surface area of the reservoir by multiplying the reference evaporation 35,36 with a coefficient 35 . Data on reference evaporation were obtained from the International Institute for Applied Systems Analysis (IIASA) 36 , which provides information for each month and on an annual basis for global land areas, excluding Antarctica, from 1961 to 1990 with a spatial resolution of 30 arc-minutes. In most cases, the hydroelectric power reservoir was smaller than the size of a grid cell in the IIASA dataset. For the reservoirs that appeared in more than one grid cell, we used the arithmetic average value of the cells. The coefficient shows the ratio of surface water evaporation to reference evaporation, and a value of 1.05 is used 35 .
To check the accuracy of the simulated reservoir evaporation (Fig. 5B) based on the reference evaporation from IIASA 36 and the coefficient 35 , we compared the calculated evaporation with the measured evaporation for the 69 reservoirs for which measured data were available. We defined the relative error (e) as the difference between the estimated and measured evaporation divided by the measured value: where e, E, and M are the relative error (%), estimated evaporation (mm yr −1 ), and measured evaporation (mm yr −1 ) for representative reservoir i in river basin j.
There are many reasons that can lead to differences between measured and simulated evaporation. These reasons include the imperfectness of simulation models, and measurement errors, among others. In this study, we found that for 90% of the reservoirs (62 out of 69), the relative error was within 30% of the measured reservoir evaporation, and for half of the reservoirs, the relative error was within 15% of the measured reservoir evaporation (see Table S7). This indicates a good estimation of the reservoir evaporation. Hence, for the reservoirs where data on measured evaporation were not available, we used the simulated evaporation.
Reservoir WF at river basin and national levels. The reservoir WF for a river basin was calculated by first summing up the reservoir WF of all representative reservoirs and then by dividing this sum by Scientific RepoRts | 5:11446 | DOi: 10.1038/srep11446 the proportion of the total reservoir area in the river basin that was accounted for by the area of the representative reservoirs: where WF is the total reservoir WF (m 3 yr −1 ) of a river basin j, N is the number of representative reservoirs in river basin j, P is the proportion of the total reservoir area in the river basin j that was accounted for by the area of the representative reservoirs in the same river basin, T is the total reservoir area (ha) in a river basin j. F and A have the same meaning and units as those in Eq. 1.
China's total reservoir WF was estimated by summing up the reservoir WFs of the 10 river basins: where W is the total reservoir WF (m 3 yr −1 ) in China and 10 is the number of total river basins in China.
Hydroelectric WF. We estimated the hydroelectric WF for 209 hydropower plants (Fig. 5C,D), which generate 53% of China's total hydroelectricity. From the Global Reservoir and Dam (GRanD) database 33 , we first selected 160 hydroelectric power plants for which hydroelectricity is the primary purpose (95) or the secondary purpose (65). For the additional 100 reservoirs (see the previous section for details), we identified 49 plants that had the production of hydroelectricity as their primary purpose. Data on the location, dam height, reservoir surface area, reservoir capacity, and electricity generation were obtained from the GRanD database 33 (see Appendix I and Appendix II).
For plants with multiple functions, we allocated the total WF among the different functions according to their economic values using an allocation coefficient η . The annual hydroelectric power WF of a specific reservoir is calculated as follows: where H is the hydroelectric power WF (m 3 yr −1 ) of reservoir i in river basin j, η is the allocation coefficient, defined as the ratio of the annual revenue generated from hydroelectric power (r, RMB yr −1 ) to the total annual revenue (R, RMB yr −1 ) generated by the hydroelectric power plant. r is calculated by multiplying the hydroelectric power that was generated (kWh yr −1 ) by the electricity price (RMB kwh −1 ). Data on electricity prices were obtained from the State Electricity Regulatory Commission 37 . We derived additional data on economic revenues from literature (see Table S7). Economic data were available for 26 hydroelectric power plants (including 13 with hydroelectric power as their main purpose), which were responsible for generating half of the hydroelectric power from the 209 representative hydroelectric power plants. We calculated η for 26 power plants that produce half of the hydroelectricity from the 209 representative plants (Table S8). For hydroelectric power plants for which no economic data were available, we used the weighted average values of η for the two categories of plants (i.e., those with hydroelectricity as their primary or secondary purpose), which were calculated based on the values for the 26 hydroelectric power plants. Here we defined two categories of hydropower plants: plants with electricity generation as their main purpose (e.g., the Three Gorges Dam in the Yangtze river, with annual electricity generation of 84.37 × 10 3 GWh) and plants with electricity generation as a secondary purpose (e.g., the Xiaolangdi Dam in the Yellow river, with annual electricity generation of 5.196 × 10 3 GWh).
For reservoirs with electricity generation as the main purpose, the weighted average η was 0.48 (Table  S8). For reservoirs with electricity generation as the secondary purpose, the weighted average η was 0.15. For the remaining 183 power plants, we use the corresponding weighted average allocation coefficients. Following Hoekstra et al. 16 , we calculated the product WF (f) of hydroelectric power, which equaled the total annual amount of water that evaporated from a reservoir that is used primarily or secondarily to generate hydroelectric power, expressed per unit of electricity generated (m 3 GJ −1 ). f was calculated by dividing the annual hydroelectric power water footprint H of a reservoir by the annual amount of electricity generated (G; GJ yr −1 ): For each basin, we calculated the proportion of the total hydropower accounted for by the representative power stations. We used this proportion, combined with the WF of the representative hydroelectric power reservoirs in the river basin, to calculate the total hydroelectric WF for the river basin. The total hydroelectric WF in China was calculated by summing up the total hydroelectric WF of the 10 major river basins: where w is the total hydroelectric WF (m 3 yr −1 ) in China, 10 is the number of total river basins in China, p is the proportion of the total hydropower accounted for by the representative power stations in river basin j, t is the annual total amount of electricity generated (GJ yr −1 ) in a river basin j.
Annual and monthly water scarcity. We quantified the annual and monthly water scarcity for the river basins with and without considering the reservoir WF. The severity of the water scarcity equaled the ratio of the WF to the water availability, which represented the difference between the natural runoff and the environmental flow requirements 38 : where S is the water scarcity indicator of river basin j, WF is the total blue water WF, Q is the actual runoff in the river basin, and the sum of Q and WF equals the natural flow 7 . r is the environmental flow requirements. WF, Q and r have a unit of m 3 ·month −1 (for monthly water scarcity calculation) or m 3 ·yr −1 (for annual water scarcity calculation).
In the above equation, we follow Hoekstra et al. 38 and define water scarcity as the ratio of blue water footprint to the blue water availability. The water footprint reflects human's water appropriation and is related to human water demand, while water availability accounts for environmental water demands by subtracting the presumed flow requirement for ecological health from the total natural runoff. Here, we follow Hoekstra et al. 38 and assume that 80% of the natural runoff should be maintained for presumed environmental flow requirements. This 80 per cent rule is proposed by water resource experts as a general precautionary guideline 38 . This means that the presumptive standard could be met as long as the human appropriation of water resources remains below 20% of the natural flow 38 .
The annual and monthly data on blue water WFs for the agricultural, industrial, and domestic sectors are available at a 30 arc-min resolution from 1996 to 2005 38 . The annual and monthly data on actual runoff at a 30 by 30 arc-min resolution were obtained from the Composite Runoff V1.0 database 39 . The annual and monthly blue water WFs of the three sectors and the actual runoff at the river basin level were calculated by aggregating the high-resolution data within a river basin. Previous calculations only considered the WF of the agriculture, industry, and domestic sectors 38 , and ignored hydroelectric power and reservoir evaporation. In our assessment of water scarcity, we calculated the results for two situations: including and excluding the reservoir WF in the total blue WF (WF) in a river basin. To calculate the monthly reservoir WF, we allocated the annual WF to each month based on the ratio of evaporation in that month to annual evaporation. These ratios were calculated using the monthly reservoir evaporation data 36 .
Water scarcity S for each basin is classified into four categories 38 : low blue water scarcity (< 100%), in which the blue water WF is less than 20% of the natural flow and does not exceed the blue water availability, river runoff is unmodified or only slightly modified, and environmental flow requirements are met; moderate blue water scarcity (100% to 150%), in which the blue water WF is between 20% and 30% of natural flow, runoff is moderately modified, and environmental flow requirements are not met; significant blue water scarcity (150% to 200%), in which the blue water WF is between 30% and 40% of natural flow, runoff is significantly modified, and environmental flow requirements are not met; and severe water scarcity (> 200%), in which the monthly blue water WF exceeds 40% of natural runoff, runoff is seriously modified, and environmental flow requirements are not met.